Is your feature request related to a problem? Please describe.
The agentic-ai module's AiFrameworkAdapter is implemented exclusively against LangChain4j. The framework's data model and release cadence block features we can otherwise reach directly through the official vendor SDKs.
Describe the solution you'd like
Replace the LangChain4j-backed framework abstraction with native implementations over the official vendor Java SDKs, preserving today's end-to-end behavior. Same provider discriminators, same element template, same hardcoded capability profile per provider. After this phase the LangChain4j framework is no longer in the default code path; its bridge factory and converters remain available as an opt-in escape hatch so users can wire a custom dev.langchain4j.model.chat.ChatModel if they need to.
This phase is the foundation for the LLM-layer ownership effort. Downstream features that depend on the new provider SPI can ship once this lands.
Describe alternatives you've considered
See the parent epic.
Is your feature request related to a problem? Please describe.
The agentic-ai module's
AiFrameworkAdapteris implemented exclusively against LangChain4j. The framework's data model and release cadence block features we can otherwise reach directly through the official vendor SDKs.Describe the solution you'd like
Replace the LangChain4j-backed framework abstraction with native implementations over the official vendor Java SDKs, preserving today's end-to-end behavior. Same provider discriminators, same element template, same hardcoded capability profile per provider. After this phase the LangChain4j framework is no longer in the default code path; its bridge factory and converters remain available as an opt-in escape hatch so users can wire a custom
dev.langchain4j.model.chat.ChatModelif they need to.This phase is the foundation for the LLM-layer ownership effort. Downstream features that depend on the new provider SPI can ship once this lands.
Describe alternatives you've considered
See the parent epic.